AFAIR to be able to use all space (block level access with VMFS) you most probably will have to divide the array in multiple <2TB LUNs. Then either create a single datastore on each LUN (recommended) or create VMFS on one and add second LUN as extent (not recommended, unless you have a very specific need for it).
In case of NFS the max size limits are set by NAS device / software.
Assuming your 2TB disks are SATA, talking about performance might be inappropriate, unless you have 2 or 3 VMs only. The first bottlenecks you are most probably going to face might be SATA HDD seek times and / or controller cache. You can try making it a bit faster by adding more spindles but anyway it's not gonna provide you with a top level speedy disk access. That's why production environments are usually based on FC/SAS/SCSI disks (they have seek times cut by ~half compared to SATA).
Of course it depends on case and specifics and sometimes SATA could be the appropriate choice (by this I'm saying that it might work as well and depends on many factors / requirements), however usually I tend not to recommend SATA based VM datastores for production environment. Only for test, backup, lab, etc.
Hope this helps.
WBR
Imants
